Skip to content

fix(session): Automatically detect and remove dangling tool calls whe…#873

Closed
liuxh613 wants to merge 2 commits intoagentscope-ai:mainfrom
liuxh613:session_restoration
Closed

fix(session): Automatically detect and remove dangling tool calls whe…#873
liuxh613 wants to merge 2 commits intoagentscope-ai:mainfrom
liuxh613:session_restoration

Conversation

@liuxh613
Copy link

@liuxh613 liuxh613 commented Mar 5, 2026

#871

Describe the bug
When restoring Agent state from a persistent session (such as a database), if the Memory contains incomplete tool calls (ToolUseBlock without a corresponding ToolResultBlock), it causes the following error:
java.lang.IllegalStateException: Cannot add messages without tool results when pending tool calls exist.
Pending IDs: [call_ba1efbfa083f4323adb4f9]

To Reproduce
Steps to reproduce the behavior:

You code
How to execute
See error
Expected behavior
The ReActAgent.loadFrom() method can include automatic cleanup logic.

Error messages
2026-03-05 12:22:57.040 INFO [] io.agentscope.core.tool.Toolkit Line:239 - Registered tool 'context_reload' in group 'ungrouped'
2026-03-05 12:22:57.040 INFO [] io.agentscope.core.memory.autocontext.AutoContextHook Line:185 - AutoContextMemory integration completed for agent: ReActAgent
Error during execution: Cannot add messages without tool results when pending tool calls exist. Pending IDs: [call_ba1efbfa083f4323adb4f9]
java.lang.IllegalStateException: Cannot add messages without tool results when pending tool calls exist. Pending IDs: [call_ba1efbfa083f4323adb4f9]
at io.agentscope.core.ReActAgent.validateAndAddToolResults(ReActAgent.java:335)
at io.agentscope.core.ReActAgent.doCall(ReActAgent.java:256)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onNext(MonoFlatMap.java:133)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onNext(FluxHide.java:137)
at reactor.core.publisher.FluxMapFuseable$MapFuseableSubscriber.onNext(FluxMapFuseable.java:130)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onNext(FluxHide.java:137)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.secondComplete(MonoFlatMap.java:246)
at reactor.core.publisher.MonoFlatMap$FlatMapInner.onNext(MonoFlatMap.java:307)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.Operators$ScalarSubscription.request(Operators.java:2565)
at reactor.core.publisher.MonoFlatMap$FlatMapInner.onSubscribe(MonoFlatMap.java:293)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.MonoJust.subscribe(MonoJust.java:56)
at reactor.core.publisher.InternalMonoOperator.subscribe(InternalMonoOperator.java:75)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onNext(MonoFlatMap.java:166)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onNext(FluxHide.java:137)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.secondComplete(MonoFlatMap.java:246)
at reactor.core.publisher.MonoFlatMap$FlatMapInner.onNext(MonoFlatMap.java:307)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.Operators$ScalarSubscription.request(Operators.java:2565)
at reactor.core.publisher.MonoFlatMap$FlatMapInner.onSubscribe(MonoFlatMap.java:293)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.MonoJust.subscribe(MonoJust.java:56)
at reactor.core.publisher.InternalMonoOperator.subscribe(InternalMonoOperator.java:75)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onNext(MonoFlatMap.java:166)
at io.agentscope.core.tracing.TracerRegistry$1.lambda$onNext$0(TracerRegistry.java:92)
at io.opentelemetry.context.Context.lambda$wrapSupplier$8(Context.java:344)
at io.agentscope.core.tracing.telemetry.TelemetryTracer.runWithContext(TelemetryTracer.java:226)
at io.agentscope.core.tracing.TracerRegistry$1.onNext(TracerRegistry.java:89)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onNext(FluxHide.java:137)
at reactor.core.publisher.Operators$ScalarSubscription.request(Operators.java:2565)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.request(FluxHide.java:152)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.request(MonoFlatMap.java:195)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.request(FluxHide.java:152)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.request(MonoFlatMap.java:195)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.request(FluxHide.java:152)
at reactor.core.publisher.FluxMapFuseable$MapFuseableSubscriber.request(FluxMapFuseable.java:172)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.request(FluxHide.java:152)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.request(MonoFlatMap.java:195)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.request(FluxHide.java:152)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.request(MonoFlatMap.java:195)
at reactor.core.publisher.Operators$MultiSubscriptionSubscriber.set(Operators.java:2361)
at reactor.core.publisher.FluxOnErrorResume$ResumeSubscriber.onSubscribe(FluxOnErrorResume.java:75)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onSubscribe(MonoFlatMap.java:118)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onSubscribe(FluxHide.java:123)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onSubscribe(MonoFlatMap.java:118)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onSubscribe(FluxHide.java:123)
at reactor.core.publisher.FluxMapFuseable$MapFuseableSubscriber.onSubscribe(FluxMapFuseable.java:96)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onSubscribe(FluxHide.java:123)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onSubscribe(MonoFlatMap.java:118)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onSubscribe(FluxHide.java:123)
at reactor.core.publisher.MonoFlatMap$FlatMapMain.onSubscribe(MonoFlatMap.java:118)
at io.agentscope.core.tracing.TracerRegistry$1.onSubscribe(TracerRegistry.java:83)
at reactor.core.publisher.FluxHide$SuppressFuseableSubscriber.onSubscribe(FluxHide.java:123)
at reactor.core.publisher.MonoJust.subscribe(MonoJust.java:56)
at reactor.core.publisher.InternalMonoOperator.subscribe(InternalMonoOperator.java:75)
at reactor.core.publisher.MonoDeferContextual.subscribe(MonoDeferContextual.java:56)
at reactor.core.publisher.InternalMonoOperator.subscribe(InternalMonoOperator.java:75)
at reactor.core.publisher.MonoUsing.subscribe(MonoUsing.java:109)
at reactor.core.publisher.Mono.subscribe(Mono.java:4569)
at reactor.core.publisher.Mono.block(Mono.java:1772)
at io.agentscope.examples.advanced.DaMengRAGFlowStudioExample.main(DaMengRAGFlowStudioExample.java:318)
Suppressed: java.lang.Exception: #block terminated with an error
at reactor.core.publisher.BlockingSingleSubscriber.blockingGet(BlockingSingleSubscriber.java:104)
at reactor.core.publisher.Mono.block(Mono.java:1773)
... 1 more

Saving final session...
Closing resources...
2026-03-05 12:22:57.116 INFO [] com.zaxxer.hikari.HikariDataSource Line:349 - HikariPool-1 - Shutdown initiated...
2026-03-05 12:22:57.118 INFO [] com.zaxxer.hikari.HikariDataSource Line:351 - HikariPool-1 - Shutdown completed.
Shutting down Studio...
✓ Done

2026-03-05 12:22:57.120 INFO [] io.agentscope.core.studio.StudioWebSocketClient Line:154 - Socket.IO disconnected from /python namespace
2026-03-05 12:23:57.137 INFO [] io.agentscope.core.model.transport.HttpTransportFactory Line:185 - Shutting down 1 managed HttpTransport(s)

Environment (please complete the following information):

AgentScope-Java Version: 1.0.10-SNAPSHOT
Java Version: 21
OS: windows11
Additional context
Add any other context about the problem here.

@liuxh613 liuxh613 requested a review from a team March 5, 2026 06:42
@gemini-code-assist
Copy link

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@codecov
Copy link

codecov bot commented Mar 5, 2026

Codecov Report

❌ Patch coverage is 81.81818% with 2 lines in your changes missing coverage. Please review.

Files with missing lines Patch % Lines
...e/src/main/java/io/agentscope/core/ReActAgent.java 81.81% 0 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!

@liuxh613 liuxh613 force-pushed the session_restoration branch from 857e491 to ee96e85 Compare March 5, 2026 07:22
@liuxh613 liuxh613 marked this pull request as draft March 5, 2026 07:39
@liuxh613 liuxh613 marked this pull request as ready for review March 5, 2026 07:39
@gemini-code-assist
Copy link

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Fixes session restoration failures for ReActAgent when persisted memory contains incomplete (dangling) tool calls, preventing IllegalStateException on the next user interaction.

Changes:

  • Add automatic cleanup on ReActAgent.loadFrom() to remove dangling tool-call state after restoring memory.
  • Add unit tests covering cleanup behavior and non-cleanup behavior on restore.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File Description
agentscope-core/src/main/java/io/agentscope/core/ReActAgent.java Runs cleanup after memory restoration to avoid pending-tool-call illegal state.
agentscope-core/src/test/java/io/agentscope/core/agent/ReActAgentTest.java Adds tests for cleanup of pending tool calls across session restore scenarios.

Comment on lines +1031 to +1046
Msg toolResultMsg =
Msg.builder()
.name("User")
.role(MsgRole.USER)
.content(
List.of(
ToolResultBlock.builder()
.id("call_complete_456")
.output(
List.of(
TextBlock.builder()
.text("Sunny, 25°C")
.build()))
.build()))
.build();
memory.addMessage(toolResultMsg);
Comment on lines +249 to +266
* session. This method removes the last assistant message if it contains tool calls without
* corresponding results, preventing IllegalStateException on the next user input.
*/
private void cleanupPendingToolCalls() {
Set<String> pendingIds = getPendingToolUseIds();
if (!pendingIds.isEmpty()) {
// Find and remove the last assistant message with pending tool calls
List<Msg> messages = memory.getMessages();
for (int i = messages.size() - 1; i >= 0; i--) {
Msg msg = messages.get(i);
if (msg.getRole() == MsgRole.ASSISTANT) {
memory.deleteMessage(i);
log.warn(
"Removed incomplete tool calls from restored session. Pending IDs: {}",
pendingIds);
break;
}
}
Set<String> pendingIds = getPendingToolUseIds();
if (!pendingIds.isEmpty()) {
// Find and remove the last assistant message with pending tool calls
List<Msg> messages = memory.getMessages();
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

getMessages() returns a filtered copy; indices might be out of sync with the internal list.

@LearningGp
Copy link
Collaborator

In my view, storing unexpected messages in the session shouldn't be handled at the framework level; instead, they should be surfaced to the developer. This might warrant further discussion.

@LearningGp LearningGp added the wait-for-response PRs that require further response label Mar 23, 2026
@LearningGp
Copy link
Collaborator

FYI (#956)

@liuxh613
Copy link
Author

FYI (#956)

thanks,This seems like a solution approach.

@liuxh613 liuxh613 closed this Mar 24, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

wait-for-response PRs that require further response

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants